As previously noted by Berg et al. (2010), six autotrophic carbon fixation mechanisms that assimilate CO₂ into cellular material have been well-documented, including the Calvin-Benson-Bassham (CBB) cycle, the reductive tricarboxylic acid (rTCA) cycle, the reductive acetyl-coenzyme A pathway, the 3-Hydroxypropionate bicycle, the hydroxypropionate-hydroxybutyrate cycle, and the dicarboxylate-hydroxybutyrate cycle. Ribulose-1,5-bisphosphate carboxylase/oxygenase (RuBisCO), one of the key enzymes associated with the first rate-limiting step for the CBB cycle, has been classified into four forms: I-IV (Zhang et al., 2016c). In previous studies, the CBB cycle was considered to be the way to fix carbon in Leptospirillum genus (Coram and Rawlings, 2002). Our present results showed that no gene encoding putative RuBisCO was identified in L. ferriphilum isolates, but non-canonical RuBisCO-like protein was predicted (Table S3). However, L. ferriphilum genomes were predicted to harbor a full suite of genes encoding canonical enzymes, which were involved in the rTCA cycle for carbon fixation. Similar results could be also found in several other reviews and papers (Levicán et al., 2008; Mi et al., 2011; Zhang et al., 2016c). In particular, genes encoding pyruvate ferredoxin oxidoreductase and 2-oxoglutarate ferredoxin oxidoreductase were predicted to exist in a common genome region (Figure S2A). In contrast, A. caldus strains were reported to harbor distinct forms of RuBisCO, and S. thermosulfidooxidans only had type I RuBisCO. For the A. caldus genomes, genes encoding type I RuBisCO were located in a gene cluster involved in carboxysome formation (Figure S2B). Notably, A. caldus isolates were predicted to harbor a carboxysome-associated carbonic anhydrase; however, no genome region related to carboxysome was identified in S. thermosulfidooxidans, despite the existence of a gene encoding the putative carbonic anhydrase.

All of the isolates were also anticipated to share a variety of genes predicted to be involved in the synthesis of lipopolysaccharides, peptidoglycan, and, potentially, exopolysaccharides (Table S3). In all of the isolates, the accumulated 3-phosphoglycerate from carbon fixation was predicted to be converted via glycolysis/gluconeogenesis to produce the major precursors (glucose-6-phosphate and glucose-1-phosphate) for the biosynthesis of cell envelope polysaccharides. In light of an earlier report using “Ferrovum” genomes (Ullrich et al., 2016a), a core set of genes that are potentially associated with amino sugar and nucleotide sugar metabolism (ko00520) were predicted in all of the investigated strains (Table S3). A variety of intermediate metabolites, such as UDP-glucose, UDP-glucuronate, UDP-galactose, and UDP-N-acetyl-glucosamine, within these metabolic processes might be potential precursors for the biosynthesis of cell envelope polysaccharides in all of the strains. Several genes that are potentially involved in the formation of precursors, such as UDP-mannose and UDP-N-acetyl-mannosamine, were only identified in the S. thermosulfidooxidans genomes, while L. ferriphilum and A. caldus were found to share the potential to produce alternative precursors, such as ADP-glucose and UDP-galacturonate. Furthermore, only the A. caldus strains appeared to catalyze GDP-mannose to GDP-4-dehydro-6-deoxy-D-mannose by GDP-mannose 4,6-dehydratase (EC 4.2.1.47).

Leptospirillum ferriphilum and Acidithiobacillus caldus were found to harbor a slightly larger enzyme repertoire related to the aforementioned biosynthesis, suggesting the production of additional macromolecules with species-specific monosaccharide compositions in comparison with S. thermosulfidooxidans. For mechanisms potentially involved in the export of cell envelope polysaccharides, the ABC-2-type polysaccharide transport system was identified in all strains, while the Lpt-type lipopolysaccharide export system was found solely in L. ferriphilum and A. caldus. Compared to the other microorganisms (Haft et al., 2006; Craig et al., 2011; Ullrich et al., 2016a), only L. ferriphilum strains shared the genes predicted to encode the PEP-CTERM exosortase system that was presumed to be involved in the export of exopolysaccharides during biofilm formation. In several earlier studies, biofilm was proposed to provide a reaction space between the sulfide surfaces and bacterial cells, thereby accelerating sulfide mineral dissolution (Watling, 2006; Rohwerder and Sand, 2007; González et al., 2013).

Acidithiobacillus caldus and Leptospirillum ferriphilum were predicted to exhibit an incomplete TCA cycle (Table S3). However, S. thermosulfidooxidans strains harbored a full suite of genes encoding the tricarboxylic acid cycle (TCA), which was thought to be essential for heterotrophic growth (Wood et al., 2004). In addition, S. thermosulfidooxidans strains exhibited the potential ability to degrade glycolic acid, as genes encoding putative glycolate oxidase and malate synthase were found in both genomes (Table S3). In particular, S. thermosulfidooxidans strains were predicted to harbor genomic regions potentially associated with glycolate oxidase (Figure S2C). A pairwise comparison of these gene clusters revealed the identical order and orientation of all of the genes, as well as the nucleotide sequence identities of up to 100%, suggesting that the gene clusters involved in glycolate oxidase might evolve from a common ancestor.

As for comparison and analysis, genome sequences of these six strains were used for the identification of carbohydrate-active enzymes (CAZymes), which represented an enormous number of enzymes that were responsible for the assembly, modification, and breakdown of oligo- and polysaccharides (Lombard et al., 2013). As shown in Table S4, the gene repertoire of potentially encoding CAZymes, including glycoside hydrolases (GHs), glycosyltransferases (GTs), polysaccharide lyases (PLs), carbohydrate esterases (CEs), auxiliary activities (AAs), and carbohydrate-binding modules (CBMs), were identified in all of the genomes. Apparently, more abundant CAZymes were identified in the S. thermosulfidooxidans genomes (156 in DX and 157 in ZJ) compared to the others (between 80 and 92), especially the GHs that are responsible for the hydrolysis of glycosidic bonds and the CEs that hydrolyze the carbohydrate esters. The most abundant GH classes in the S. thermosulfidooxidans isolates were strongly biased toward the catabolism of oligo- or polysaccharides, such as chitinase (GH18), cellobiohydrolase (GH74), glucoamylase (GH15), and α-amylase (GH13).

Similar to an earlier study (Yelton et al., 2013), another potential organic carbon source for the S. thermosulfidooxidans strains might be the lipids from the lysed cells. They were predicted to harbor a full set of putative proteins involved in fatty acid degradation (Table S3), including long-chain acyl-CoA synthetase, acyl-CoA dehydrogenase, enoyl-CoA hydratase, and acetyl-CoA acyltransferase. Additionally, lactate permease and lactate utilization of protein B/C was identified in the S. thermosulfidooxidans isolates (Table S3), suggesting their heterotrophic growth on lactate.

Acidithiobacillus caldus strains harbor the complete genes required for dissimilatory nitrate reduction, which were observed as NarGHJI and nirBD operons (Table S3). Thus, it was hypothesized that the A. caldus strains in this study might utilize nitrate and nitrite as electron acceptors under anaerobic environmental conditions. As for assimilatory nitrate reduction, A. caldus was predicted to harbor narA potentially as an encoding assimilatory nitrate reductase catalytic subunit; however, there was no NasB subunit, which was reported to transfer electrons from NADH to nitrate (Lin and Stewart, 1997), thereby making the electron donor unclear. Additionally, no nirA gene was identified in the A. caldus genomes. In contrast, S. thermosulfidooxidans share the set of genes that encode assimilatory nitrate reduction. We identified the genes encoding the assimilatory nitrate reductase catalytic subunit (NasA) and ferredoxin-nitrite reductase (NirA). Although putative genes potentially encoding dissimilatory nitrite reductase small subunit (NirD) and ferredoxin-nitrite reductase (NirA) were found in the L. ferriphilum genomes, their functional roles need to be studied further. Additionally, all of the strains share the potential to utilize ammonium via the Amt family transporter (Table S3), and to assimilate it into the central metabolic pathways via glutamine synthetase or glutamate synthase. Given that dissimilatory nitrate reduction and assimilatory nitrate reduction were absent in L. ferriphilum, they might utilize ammonium as the sole nitrogen source. In addition, S. thermosulfidooxidans was the only species that contains gene cluster encoding urease subunits (UreABC) and accessory proteins (UreDEFG; Table S3). Accordingly, it seems that S. thermosulfidooxidans could utilize extracellular urea, although there is still no direct evidence showing the existence of these compounds in the bioleaching heaps.

Both L. ferriphilum and S. thermosulfidooxidans are demonstrated iron oxidizers (Guo et al., 2014; Justice et al., 2014; Zhang et al., 2016c). These two species were predicted to harbor membrane-associated c-type cytochromes that are potentially involved in iron oxidation. In L. ferriphilum, the electrons might be transferred from an outer-membrane cytochrome c (Cyc₅₇₂) to a periplasmic cytochrome c (Cyt₅₇₉), and then across periplasmic cytochromes c either reducing oxygen via cbb₃-type terminal oxidase (downhill), or reducing NAD⁺ passing through the bc₁ complex, quinone pool, and NADH dehydrogenase (uphill, Bonnefoy and Holmes, 2012). Similar to Cyt₅₇₉ in L. ferriphilum, the periplasmic sulfocyanins identified in S. thermosulfidooxidans (Table S3) were postulated to be the branch point at which electrons derived from membrane c-type cytochromes might be channeled either downhill or uphill. As for the uphill pathway, the bc₁ complex in Leptospirillum spp. was proposed to utilize the proton motive force (PMF) to push electrons uphill against the thermodynamically unfavorable gradient (Bonnefoy and Holmes, 2012). Herein, the outflux of protons catalyzed by the cbb₃-type terminal oxidase contributes to the generation of the PMF during electron transfer to oxygen via a downhill pathway. However, genes encoding the putative bc₁ complex were absent in S. thermosulfidooxidans. It seems that this species may not require reverse electron transfer to generate reduction equivalents (e.g., NADH and NADPH), but depend on the oxidation of organic compounds, such as TCA cycle, oxidative pentose phosphate pathway, to produce these compounds, since it was reported to be mixotroph capable of the assimilation of both inorganic and organic carbon compounds. In addition, A. caldus and S. thermosulfidooxidans analyzed here were replete with metabolic enzymes related to sulfur oxidation (Table S3), and they have the ability to oxidize minerals, which produce sulfuric acid (Figure 4). Their activities facilitate the dissolution of base metals in certain acid-soluble sulfide minerals, such as chalcocite, via an oxidative route (Johnson and Hallberg, 2003). Oddly, gene encoding putative thiosulfate:quinone oxidoreductase (TQO) was absent in both A. caldus strains, while the corresponding gene was identified in the homologous strains such as SM-1 and ATCC 51756 (Zhang et al., 2016b). In addition, gene encoding sulfate adenylyltransferase (SAT) was not found in S. thermosulfidooxidans genomes analyzed in this study, however, the homologous strain ST was reported to harbor the sat gene (Guo et al., 2014; Zhang et al., 2017). In A. caldus (Chen et al., 2012), electrons from TQO, sulfide quinone reductase, sulfur oxidizing protein, and heterodisulfide reductase were mediated by the quinol pool, and then (i) either were transferred to oxygen via bd-type or bo₃-type terminal oxidases, (ii) or were transferred to NADH complex I (Table S3). Similarly, genes encoding putative bd-type terminal oxidase and NADH complex I were identified in S. thermosulfidooxidans genomes.

An earlier study revealed that Acidithiobacillus spp. and/or Leptospirillum spp. were observed to be the major carbon fixers in acid mine drainage (AMD) communities (Chen et al., 2015). Previous studies showed detailed evidence supporting the fixation of carbon dioxide via the classical CBB cycle in A. caldus (You et al., 2011; Zhang et al., 2016b) and S. thermosulfidooxidans (Guo et al., 2014; Justice et al., 2014; Zhang et al., 2017). Apart from autotrophic growth, S. thermosulfidooxidans isolates have been reported to grow heterotrophically via the catabolism of various organic carbon compounds (Guo et al., 2014; Justice et al., 2014). Unlike S. thermosulfidooxidans, A. caldus was predicted to elevate the concentration of carbon dioxide near RuBisCO by a carboxysome-associated carbonic anhydrase, which was responsible for the conversion of accumulated cytosolic bicarbonate into CO₂ (Zhang et al., 2016e). Thus, carboxysome within A. caldus strains might be a critical component that contributes to their ability to grow autotrophically through the efficient utilization of scarce carbon availability from acidic environments.

Certain organic compounds, such as cell exudates or lysates that originated from the autotrophic primary producers might be utilized by heterotrophic prokaryotes (Johnson and Hallberg, 2003). Glycolic acid (C00160) was reported to be an exudate originating from autotrophic acidophiles, such as Leptospirillum (Justice et al., 2014). In contrast to Sulfobacillus acidophilus (Justice et al., 2014), no genes encoding putative isocitrate lyase (EC 4.1.3.1) were identified in S. thermosulfidooxidans. Accordingly, we surmise that S. thermosulfidooxidans might not have the ability to assimilate isocitrate to malate via glyoxylate bypass, a pathway closely related to the TCA cycle (Yelton et al., 2013). In addition, a gene cluster involved in glycolate oxidase was predicted in S. thermosulfidooxidans. A putative transposase identified in its downstream region suggested that this cluster might be introduced by HGT, thereby recruiting a novel ability to degrade the organic carbon (Figure S2C). Taken together, we presented a potential route of glycolic acid assimilation that enables S. thermosulfidooxidans to utilize organic acid as an alternative carbon source (Figure S3). In this pathway, glycolate oxidase coupled with malate synthase may degrade glycolic acid to produce malate, and the latter then enters the conventional TCA cycle. Chemolithotrophic acidophiles were sensitive to organic acids in their habitats (Johnson, 1998); therefore, it might be favorable for chemoautotrophs to form a mutualistic relationship with other heterotrophic and/or mixotrophic acidophiles because the latter could effectively detoxify the environments by degrading organic compounds (Johnson and Hallberg, 2003).

In addition, CDS assigned to chitinases (GH18) were found in S. thermosulfidooxidans. The GH18 family contained chitinolytic enzymes involved in chitin degradation, and chitin is known to be a substantial component of fungal cell walls. An intriguing explanation for the abundant chitinolytic enzymes in Trichoderma is that chitinases are regarded as an integral part of the mycoparasitic lifestyle via lysis of the prey's cell walls (Kubicek et al., 2011). In our study, we could not exclude the possibility that putative chitinases only present in S. thermosulfidooxidans might endow them with an additional ability to degrade the lysates of fungi in bioleaching systems. This scenario might be reasonable as fungi were found in several acidic environments, such as AMD ecosystems (Mosier et al., 2016), although their functional roles in bioleaching heaps remain elusive (Cárdenas et al., 2016).

Like other microorganisms in AMD that are characterized by scarce nitrogen availability (Parro et al., 2007; Chen et al., 2015), the habitants in the bioleaching systems may share nitrogen assimilation strategies to cope with the nitrogen-limited conditions. In general, microorganisms commonly utilize atmospheric nitrogen (N₂), ammonium, nitrite, and nitrate as their inorganic nitrogen sources (Arrigo, 2005). As a member of the Acidithiobacillus genus, A. ferrooxidans has been predicted to harbor nitrogenase, which is responsible for N₂ fixation (Levicán et al., 2008; Zhang et al., 2016e). Likewise, nitrogen fixation was also reported in the Leptospirillum groups, except for Group II (Goltsman et al., 2013). Under the low nitrogen level, the diazotrophic lifestyle of these bacteria in the microbial communities might be a striking feature to maintain the concentration of nitrogen availability in the habitat environments.

Although the three species analyzed in this study do not have the ability to fix atmospheric nitrogen, they were predicted to utilize environmental ammonium, nitrite, and even nitrate as an alternative nitrogen source. L. ferriphilum was predicted to merely harbor the ability of ammonium uptake, but the other two species could utilize other forms of nitrogen, such as nitrate or nitrite as the alternative nitrogen source, thereby, to some extent, avoiding competitive exclusion. Notably, microbial ureases have been reported to hydrolyze environmental urea to produce ammonia and bicarbonate (Mobley and Hausinger, 1989; Ullrich et al., 2016a,b), and ammonia at circumneutral cytoplasmic pH is present as ammonium (Ullrich et al., 2016b). Thus, the S. thermosulfidooxidans strains appear to be able to incorporate cytoplasmic ammonium derived from the hydrolysis of urea as a nitrogen source (Figure 4), and utilize the released bicarbonate as an alternative carbon source via carbonic anhydrase.

The models for potential iron-sulfur cycling were well-studied in acidic and high-sulfur environments (Johnson, 1998; Baker and Banfield, 2003; Méndez-García et al., 2015; Chen et al., 2016; Zhang et al., 2016e). In these settings, ferrous iron oxidation implemented by iron-oxidizing bacteria, such as L. ferriphilum and S. thermosulfidooxidans, is the primary biochemical transformation of iron. In addition, the oxidation of sulfide minerals, elemental sulfur, and tetrathionate has been well-documented in A. caldus and S. thermosulfidooxidans (Mangold et al., 2011; Chen et al., 2012; Guo et al., 2014; Zhang et al., 2016b). Fe(III) under anoxic conditions could accept electrons originating from inorganic (e.g., sulfur and/or hydrogen) and/or organic donors (e.g., glucose and/or glycerol), which occurs in the case of chemolithotrophic and heterotrophic acidophiles, respectively (Méndez-García et al., 2015). When exposed to moisture and air, sulfur-containing mineral ores oxidize spontaneously in situations where molecular oxygen or Fe(III) act as an oxidant (Johnson and Hallberg, 2003). The formation of acidic environments in turn provides advantages to stabilizing Fe(II), even in the presence of atmospheric oxygen (Bonnefoy and Holmes, 2012); this is an opportunity for iron-oxidizing microorganisms inhabiting these settings to utilize Fe(II) as a source of energy.

Acidithiobacillus caldus and Sulfobacillus thermosulfidooxidans strains used in this study were predicted to lack TQO and SAT encoding genes, respectively. One plausible explanation for this issue was that missing genes might be artifacts of assembly or gene calling procedure. On the other hand, we could not exclude the possibility that gene tqo in homologous strains of A. caldus (SM-1 and ATCC 51756) and sat in homologous strains of S. thermosulfidooxidans (ST) might be introduced by HGT. With respect to ferric iron [Fe(III)] reduction, previous studies have shown that Sulfobacillus spp. harbored the ability to reduce ferric iron (Melamud et al., 2003; Bogdanova et al., 2006; Johnson et al., 2008). Although Fe(III) reduction activities under oxygen-limiting conditions were observed in acidophilic microorganisms, the determination of a precise enzymatic system is still an important goal for future studies.

Comparative genomics of three typical co-occurring acidophiles in bioleaching heaps yields valuable insights into the differences in gene content, metabolic profiles, and potential functional roles. In short, it could present a scenario whereby certain essential functions are partitioned in co-occurring members of microbial communities, and individual members harboring different metabolic functions generate not merely commensalistic but mutualistic interactions. Thus, it might provide a selective advantage for the efficient utilization of limited resources in their habitats in which they benefit from each other to maintain their respective lifestyle.